Inspiration

A majority of critical data is streaming data. One such example is the performance of factory machines over time. We wanted to devise a method in order to analyze these machines for potential performance errors and to visualize their performance over time. This would tremendously help cut business costs, as specialized technicians would be able to prevent these performance errors before they occur, in order to save the company resources.

What it does

Our app is a platform for predictive analysis on real-time machine performance data from Black & Decker. Hence, its primary purpose is to anticipate whether a machine would fail based on the given data. A secondary objective was to build an IoT dashboard to visualize results of exploratory data analysis (e.g. summary statistics).

How we built it

In order to perform predictive analysis, we implemented an anomaly detection algorithm that takes the temporal nature of the data into consideration. We implemented the system on top of the Flask framework, with a React front-end. In addition, we used Firebase as a central data repository and hosted the app on Google's App Engine. We also used the Twilio API in order to implement text-message alerts.

Challenges we ran into

The dataset we were given to work with was very small and not representative of the entire problem domain. Hence, we had to design a custom algorithm in order to generate sufficient data. Also, none of us had experience working with time series models or anomaly detection, so we had to learn a lot along the way.

Accomplishments that we're proud of

We're most proud of the fact that we were able to collaborate independently on different features of the system (e.g. algorithms, back-end, front-end) and still managed to fuse everything together into the final product. Also, we're proud that we were able to tackle such a challenging problem without any obvious solutions within such a short timespan.

What we learned

We learned a lot about the IoT framework and predictive analysis on streaming data (e.g. online learning, time series, etc.). In addition, we learned about the importance of clean data for both visualization and predictive analysis. Last but not least, we learned that teamwork and collaboration are major keys to tackling challenging and unfamiliar problems.

What's next for IoT Streaming and Analytics Dashboard

To infinity and beyond!

Share this project:
×

Updates